Benefits of Pod Dimensioning With Best-Effort Resources in Bare Metal Cloud Native Deployments

Container orchestration platforms automatically adjust resources to evolving traffic conditions. However, these scaling mechanisms are reactive and may lead to service degradation. Traditionally, resource dimensioning has been performed considering guaranteed (or request) resources. Recently, container orchestration platforms included the possibility of allocating idle (or limit) resources for a short time in a best-effort fashion. This letter analyzes the potential of using limit resources as a way to mitigate degradation while reducing the number of allocated request resources. Results show that a 25% CPU reduction can be achieved by relying on limit resources.